term2gif

Record a terminal session and save it as a GIF, in a single binary (term2gif) for GNU/Linux and Windows, with no external tools to install.

term2gif [OPTIONS] <OUTPUT.gif>

term2gif demo.gif                                    # record a shell session
term2gif demo.gif -c 'ls -la'                        # record one command
term2gif demo.gif --theme monokai --speed 1.5 --idle-time-limit 2
term2gif demo.gif --keep-cast demo.cast              # also keep the .cast file
term2gif demo.gif --from-cast demo.cast              # render an existing recording
term2gif demo.gif --bitmap-font fixed                # look like xterm's default font

Inside xterm (or uxterm), the GIF uses xterm's font automatically (see below).

End the recording by exiting the shell (exit / Ctrl-D).

Options

Recording:

Rendering and timing:

How it works

• Recording: the shell or command runs in a PTY via portable-pty (Unix PTY, Windows ConPTY). Its output is shown in your terminal and written to an asciicast v2 file in a temporary location, which is deleted afterwards unless you use --keep-cast.
• Rendering: the recording is replayed through a virtual terminal (avt, the one asciinema uses), and each frame is drawn with the built-in DejaVu Sans Mono font (swash for rasterizing). Box drawing and block characters are drawn directly, so lines connect at any line height.
• X11 bitmap fonts (--bitmap-font): the classic xterm fonts in PCF format, found through the fonts.dir / fonts.alias files of the X11 font directories (on Debian: package xfonts-base). As in xterm (uxterm), the ISO 10646 version of the font is used, with its bold variant (or overstriking, if there is none) and a double-width font of the same height for CJK characters (e.g. 12x13ja for fixed). Characters the bitmap fonts lack are drawn with the other fonts, scaled to the cell size. On Windows, pass a font file.
• xterm font detection: when term2gif runs inside xterm (recognized by its reply to the XTVERSION query, or an xterm parent process) and no font option is given, it uses the font xterm uses. It asks xterm for its current font (OSC 50; needs XTerm*allowFontOps: true, and then also follows changes from the font menu), and otherwise takes it from the xterm command line (-fn, -fa, -fs, -class UXTerm), the X resources (xrdb -query: font, faceName, faceSize, renderFont) or xterm's default fixed. TrueType fonts (faceName) are used at xterm's size, with the font's own line height as in xterm.
• Unicode: wide characters (CJK, emoji) take two cells. Characters missing from the font are taken from any installed font (fontdb); color emoji fonts (e.g. Noto Color Emoji, Segoe UI Emoji) are preferred for emoji.
• Resizing: window resizes during recording are forwarded to the program and recorded. A GIF has one fixed size, so it is as large as the largest terminal size during the session; while the terminal is smaller, the unused area shows the background color.
• GIF encoding: gif. Each frame stores only the changed area, and terminal colors are kept exact.

Install

• Prebuilt binary: download the archive for GNU/Linux (x86_64, static, so it runs on any distribution) or Windows (x86_64) from the releases and put term2gif on your PATH.
• With uv (PyPI package term2gif-rs, which contains the same binary): uv tool install term2gif-rs, or without installing: uvx --from term2gif-rs term2gif demo.gif.
• With cargo: cargo binstall term2gif (downloads the prebuilt binary) or cargo install --locked term2gif (builds it, needs Rust 1.86 or newer).

Windows needs Windows 10 version 1809 or newer (ConPTY).

Build

With Rust 1.86 or newer, straight from GitHub:

cargo install --locked --git https://github.com/sbellon/term2gif

or from a local checkout:

cargo install --locked --path .

Either installs term2gif into ~/.cargo/bin. Or build it with cargo build --release and put target/release/term2gif (term2gif.exe on Windows) on your PATH yourself.

Checks

The typical Rust flow must stay clean (clippy runs with the pedantic group enabled via [lints.clippy] in Cargo.toml):

cargo fmt --check
cargo check --all-targets --all-features
cargo clippy --all-targets --all-features -- -D warnings
cargo test --all-features
cargo doc --no-deps --all-features

plus a clippy pass for the other platform's code via a cross target (x86_64-pc-windows-msvc from GNU/Linux, x86_64-unknown-linux-musl from Windows), e.g.:

cargo clippy --all-targets --all-features --target x86_64-pc-windows-msvc -- -D warnings

A pre-commit hook runs all of it automatically; enable it once per clone with:

git config core.hooksPath .githooks

Limitations

• Keystrokes are not captured in the recording (output only).
• Combining characters (e.g. a separate accent after a letter) are dropped by the virtual terminal; precomposed characters like ä work.
• No ligatures or right-to-left text shaping; each character is drawn in its own cell, as in a terminal.

License

MIT, see LICENSE. The bundled DejaVu Sans Mono fonts are under the Bitstream Vera license, see fonts/DejaVu-LICENSE.txt. Dependencies are under MIT, Apache-2.0 or similar permissive licenses.